Udi, my interpretation of Greg's theory....
For example, an extremely LTL gyro with no HS experiences a sudden power failure at high airspeed. It nose drops (a stable response), but, because of low rotor damping and some moderate value of frame MOI, it just keeps going. The rotor loses AOA and therefore thrust. It becomes too weak to arrest the pitchover. The gyro simply goes upside down in a "pure bunt."
I'm not convinced that this is possible with real values of frame MOI, thrustline location and rotor damping, but you can at least make up numbers that might work. I suppose we need a name for this kind of scenario -- diabolis ex machina? One might call it dynamic instability if that name weren't already taken.